On Wed, Jul 23, 2014 at 6:10 AM, Neil Roberts <[email protected]> wrote: > Ilia Mirkin <[email protected]> writes: > >> I may be misunderstanding, but it seems like this test only tests the >> RGBA_UNORM format. How hard would it be to test all 4 formats? If it's >> a pain to do so in the way you're doing here, you could have a more >> basic test that just ensures that a texture that was compressed by the >> library can be used for texturing (and returns the data you'd expect >> it to). > > Yes, it only tests the decompressor for one of the normalized formats. I > should have made that more clear in the commit message and in the > comment in the source code. You're right there needs to be a bunch more > tests to test the half-float formats and the compressor. > > It might be a bit tricky to check that you get the data you expect from > the compressor because we would want quite a lot of leeway for the > compressor to give an estimation of the image. Maybe it would be enough > to just compress some simple images that don't have more than two > colours per block because these should probably compress quite reliably.
Is there a way to read back the image once it's compressed that ensures that software would be used to unpack it? If so, just compare that using the texture in a shader == reading it back post compression. Anyways, however you want to test is fine with me. My main point is that it doesn't _have_ to be some huge complex thing, as long as there's some verification that texturing from a shader/blits isn't totally broken. But perhaps there's no easy way to do that. -ilia _______________________________________________ Piglit mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/piglit
